• Developed and deployed Profé, a Spanish-language LLM chatbot that assists users with tutorials, teaches financial concepts, and promotes good financial habits.

• Led real-time fraud detection implementation for card transactions with sub-1-second response times, enabling point-of-sale declines based on data-driven insights to protect cardholders.

• Built and deployed the backend service for a banking platform from scratch as part of a 4-person team in 6 months, implementing REST APIs, compliance logic, transaction recording, notifications, document uploads, and testing using Java, Spring, and Postgres.

• Contributed to critical infrastructure tasks, including AWS configuration, firewall implementation, CI/CD pipeline development, third-party API integration, and production/hotfix deployments.

• Owned data pipeline maintenance, API integration, dashboard development, and data analytics; served as the data subject matter expert for executives and cross-functional teams.

• Led development of BillBert, a custom deep learning NLP and CV algorithm enabling sub-10ms search across 2M+ documents, driving $700M/year in low-risk payments and $15M/year in revenue.

• Delivered a patented Python computer vision model (US12046063B2) with over 90% accuracy for extracting line items, enabling real-time predictions for 1.2M+ documents monthly and surpassing competitor performance.

• Developed and productionized machine learning models to enhance payment processing and user predictions, including void propensity ranking for $2B/year workflows and personalized due date forecasting with a 14% accuracy improvement.

• Won multiple internal awards, including First Place Overall in Hackathon for creating an LLM-powered financial insights tool and Judge’s Choice Award for an NLP model predicting customer service outcomes.

• Revamped repeat-customer purchase propensity model using 20M+ data points of user data to enable improved selection of customers to receive direct mail, driving revenue growth and reducing marketing spend